Re: Two small.... + Faq # 17

["Kristian K. Nielsen" <ap...@ctav.com>]

Sorry never saw the PR#2366, also I did the patch last weekend and at that
point there wasn't any patches ;-)

Anyway...I've read your answer to FAQ question #17 ("How can I have my
script output parsed?........This is a feature The Apache Group gopes to add
in the next major release"):

Sorry, I'm new to this list and I was just wondering what the current state
on this manner is.
If nobody has started the preparation yet, I would like to help.

If not, which method does The Apache Group have in mind (first about
handling)?:
1. The server should rehandle all output from the script based on the
Content-Type, e.x. the script returns the old "text/x-server-parsed-html",
in the configuration file there should be two new Options, say:
    ScriptType text/html text/x-server-parsed-html
    ScriptHandler server-parsed text/x-server-parsed-html
2. Same as above but handled inside the script using either a speciel unique
Status-code or a new script header, e.x. Content-Handle.

How to implement it into the source
1. Both of above methods could be made rather easily (and messy) by
buffering the entire output from the script in either a large memory block
or a temporary file, simply by rewriting parts of mod_cgi and parts of the
main-sourcecode.
2. A more throught method would be to rewrite the buffering system, so input
can be handled directly from mod_cgi to modules like mod_include.

Yes, I don't know much about what I'm taking about at this point, but I
will - doing next week I hope :-)

Kristian

-----Original Message-----
From: Dean Gaudet <dg...@arctic.org>
To: Kristian K. Nielsen <ap...@ctav.com>
Cc: new-httpd@apache.org <ne...@apache.org>
Date: 10 June 1998 06:12
Subject: Re: Two small mistakes in the release of Apache 1.3


>We're talking about releasing 1.3.1 in a little over a week.  You can find
>the specific patch in PR#2366 at http://bugs.apache.org/index/full/2366.
>
>Dean
>
>On Wed, 10 Jun 1998, Kristian K. Nielsen wrote:
>
>>
>> >
>> >I nailed that one already... thanks for pointing it out though.  I
changed
>> >the cookie format to actually include the FQDN... it seems silly to not
>> >use the FQDN... but maybe someone will explain why the unqualified name
is
>> >good.
>> >
>>
>>
>> Then when can we except the officiel patch to be released?
>>
>> Kristian
>>
>>
>
>

Re: Two small.... + Faq # 17

[Dean Gaudet <dg...@arctic.org>]

No there won't be any layering in 1.x.

Dean

On Thu, 11 Jun 1998, Chia-liang Kao wrote:

> Does this mean that the protocol abstraction and this feature won't be in 
> the future 1.x tree?
> 
> Recently I have a job on this, what I have in mind is something kinda 
> brute force: change the outgoing fd in r->connection->client to a 
> external program(in fixup?), and pass the real outgoing file descriptor to 
> that filter.  So we can do things like SSL without modifying the source, 
> but cost deduced.
> 
> Another approach is having some functions registered to BUFF as pre-write
> filter functions, but the sequence of multiple filtering functions is
> something hard to handle. I'd call this ugly layering(stacking?).. :P
> 
> 
> CLK
> 
> On Wed, Jun 10, 1998 at 12:34:53AM -0700, Dean Gaudet wrote:
> > On Wed, 10 Jun 1998, Kristian K. Nielsen wrote:
> > 
> > > Anyway...I've read your answer to FAQ question #17 ("How can I have my
> > > script output parsed?........This is a feature The Apache Group gopes to add
> > > in the next major release"):
> > > 
> > > Sorry, I'm new to this list and I was just wondering what the current state
> > > on this manner is.
> > > If nobody has started the preparation yet, I would like to help.
> > 
> > If you want a kludge for 1.x it's not hard -- you just need to hardwire
> > mod_cgi and mod_include together.  mod_include only cares that it get a
> > FILE *. 
> > 
> > The full solution requires I/O layering, and that won't happen until 2.0. 
> > 
> > We should be more proactive about producing a set of 2.0 design goals so
> > that folks can begin working.  I know several folks that have offered to
> > help and I just don't know where to point them.
> > 
> > Dean
> 

Re: Two small.... + Faq # 17

[Chia-liang Kao <cl...@oneway.net>]

Does this mean that the protocol abstraction and this feature won't be in 
the future 1.x tree?

Recently I have a job on this, what I have in mind is something kinda 
brute force: change the outgoing fd in r->connection->client to a 
external program(in fixup?), and pass the real outgoing file descriptor to 
that filter.  So we can do things like SSL without modifying the source, 
but cost deduced.

Another approach is having some functions registered to BUFF as pre-write
filter functions, but the sequence of multiple filtering functions is
something hard to handle. I'd call this ugly layering(stacking?).. :P


CLK

On Wed, Jun 10, 1998 at 12:34:53AM -0700, Dean Gaudet wrote:
> On Wed, 10 Jun 1998, Kristian K. Nielsen wrote:
> 
> > Anyway...I've read your answer to FAQ question #17 ("How can I have my
> > script output parsed?........This is a feature The Apache Group gopes to add
> > in the next major release"):
> > 
> > Sorry, I'm new to this list and I was just wondering what the current state
> > on this manner is.
> > If nobody has started the preparation yet, I would like to help.
> 
> If you want a kludge for 1.x it's not hard -- you just need to hardwire
> mod_cgi and mod_include together.  mod_include only cares that it get a
> FILE *. 
> 
> The full solution requires I/O layering, and that won't happen until 2.0. 
> 
> We should be more proactive about producing a set of 2.0 design goals so
> that folks can begin working.  I know several folks that have offered to
> help and I just don't know where to point them.
> 
> Dean

Re: Two small.... + Faq # 17

[Dean Gaudet <dg...@arctic.org>]

On Wed, 10 Jun 1998, Kristian K. Nielsen wrote:

> Anyway...I've read your answer to FAQ question #17 ("How can I have my
> script output parsed?........This is a feature The Apache Group gopes to add
> in the next major release"):
> 
> Sorry, I'm new to this list and I was just wondering what the current state
> on this manner is.
> If nobody has started the preparation yet, I would like to help.

If you want a kludge for 1.x it's not hard -- you just need to hardwire
mod_cgi and mod_include together.  mod_include only cares that it get a
FILE *. 

The full solution requires I/O layering, and that won't happen until 2.0. 

We should be more proactive about producing a set of 2.0 design goals so
that folks can begin working.  I know several folks that have offered to
help and I just don't know where to point them.

Dean